The issue mentioned in that post - the fact that this query generates 6 sql queries instead of 2 {allPosts(first:5){title}} is a limitation of the implementat

The issue mentioned in that post - the fact that this query generates 6 sql queries instead of 2 {allPosts(first:5){title}} is a limitation of the implementation and not inherent to the architecture. We spent a lot of time optimising these cases for graph.cool and the result is that we actually have a more scalable system by issuing a few simple sql queries instead of one giant join. It does open the door for the result to be internally inconsistent, but that is no different from querying a bunch of cached rest endpoints.